npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

epsilon-greedy

v0.0.1

Published

Abstract Epsilon Greedy algorithm.

Downloads

3

Readme

epsilon-greedy

Abstract Epsilon Greedy algorithm.

This is an abstract implementation for the Epsilon Greedy algorithm.

It's an abstract version, so you can use this module for any kind of split-test you'd like.

Usage:

var epsilonGreedy = EpsilonGreedy(?opts);

First you need create an instance of the EpsilonGreedy object.

options:

  • pathToTrials - (Optional|default=trials). Path to your trials property inside your variant object.
  • pathToRewards - (Optional|default=rewards). Path to your rewards property inside your variant object.
  • handlers - (Optional array of handlers). Read more about handlers bellow.

Handlers

Handlers are the functions that decide which variant will win.

Here is how you define the handlers array:

var epsilonGreedy = EpsilonGreedy({
  handlers: [
    {epsilon: 0.1, handler: function someHandler (variant) {//some code}},
    {epsilon: 0.1, handler: function anotherHandler (variant) {//some code}},
    function finalHandler(variant) { //some code}
  ]
});

The way it works, is that it'll pick someHandler 10% of the tests. 90% of the tests would be handled by other handlers. Which means, that anotherHandler would be picked 10% * 90% of the tests, which is 9%.

finalHandler does not have an epsilon assigned to it. Whenever you're not specifying an epsilon, the epsilon would be equal to 1.

So eventually we could write the last line as:

{epsilon: 1, handler: function finalHandler(variant) { //some code}}

Predefined Handlers

epsilon-greedy comes out of the box with some ready to go handlers.

EpsilonGreedy.handlers.randomExploration

This handler will pick a random variant.

EpsilonGreedy.handlers.trialsBellowAverageExploration

In the original EpsilonGreedy algorithm, it's recommended to run 1,000 tests on each variant prior using this algorithm.

It might be a problem in 2 cases:

  • When you're constantly adding new variants.
  • When you need way more trials for a reward.

This handler will calculate the amount of trials needed for 1 reward and it'll pick the the first variant that has no rewards and it's amount of trials is bellow the average amount. The variants would be sorted in ascending order based on amount of trials.

In case all variants has rewards, this handler will be ignored and it's epsilon would pass to the next handler.

EpsilonGreedy.handlers.bestVariantExploitation

This handler will pick the variant with the highest success rate.

EpsilonGreedy.handlers.numberedExploiter(index)

In some cases you may want to give some test share for the 2nd best variant or the 3rd best variant as well.

This handler will pick a numbered variant based on success rate.

index is a number between 0 and variants.length-1.

If you specify a number that is out of the scope, the index would be 0 (aka the best variant).

Handler priority

If the handler can't handle the current test, EpsilonGreedy will skip the handler and will split it's epsilon value to the rest of the handlers.

An example for a use case it might happen would be the case of trialsBellowAverageExploration. This handler would be ignored in case all the variants has at least one reward or all the variants has passed the minimum amount of trials for 1 reward.

Usage:

{epsilon: 0.1, handler: EpsilonGreedy.handlers.numberedExploiter(1) // Would pick the 2nd best

Developing your own handlers

A handler is a basic js function that gets the variants as it's argument.

It should return a variant or undefined. If it returns undefined, this handler will be skipped and it's epsilon will be shared by the remaining handlers in the pipeline.

Example:

function randomExploration(variants) {
  return _.shuffle(variants)[0];
}

Default Handlers

The default handler setup is the most popular way of using this algorithm:

var epsilonGreedy = EpsilonGreedy({
  handlers: [
    {epsilon: 0.1, handler: randomExploration},
    bestVariantExploitation
  ]
});

My favorite way of structuring handlers

var epsilonGreedy = EpsilonGreedy({
  handlers: [
    {epsilon: 0.1, handler: randomExploration},
    {epsilon: 0.1, handler: trialsBellowAverageExploration},
    bestVariantExploitation
  ]
});

var winner = epsilonGreedy(variants)

Finally, after everything is set, we wanna get our winning variant for the current test.

variants should be an array of json objects. Each variant should have a property that contains the amount of trials and the amount of rewards.

You can specify the name of those properties in EpsilonGreedy options.

Example

var EpsilonGreedy = require('epsilon-greedy');

var epsilonGreedy = EpsilonGreedy({
  handlers: [
    {epsilon: 0.1, handler: EpsilonGreedy.handlers.randomExploration},
    {epsilon: 0.1, handler: EpsilonGreedy.handlers.trialsBellowAverageExploration},
    EpsilonGreedy.handlers.bestVariantExploitation
  ]
});

console.log(epsilonGreedy([
  {trials: 100, rewards: 1},
  {trials: 70, rewards: 2},
  {trials: 80, rewards: 0},
  {trials: 32, rewards: 0},
  {trials: 20, rewards: 0}
]));

install

With npm do:

npm install epsilon-greedy

license

MIT